基于贝叶斯原理的大洋金枪鱼渔场速预报模型研究

Tuna fishing grounds prediction model based on Bayes probability

高度洄游的大洋性金枪鱼类(Scombridae)是世界远洋渔业的重要捕捞对象,中国金枪鱼生产仍处于初期发展阶段,因此研究和预测金枪鱼渔场具有重要的现实意义。本研究利用美国NASA提供的卫星遥感反演海表温度(SST)三级数据产品和太平洋共同体秘书处(SPC)提供的有关国际金枪鱼历史捕捞产量资料,分析金枪鱼同SST等海洋渔场环境要素之间的统计关系,建立了金枪鱼渔场的贝叶斯概率预报模型。通过对历史数据进行模型回报试验,结果表明太平洋鲣鱼渔场综合预报的准确性达到70%以上,对渔业捕捞生产具有一定的指导意义。

Highly migration tuna is one of the most important fishing objects of high sea fisheries of the world. At present, the tuna fishery of China is at the primary development period, and it is practically significant to engage in forecasting and studying on tuna fishing grounds. Based on three-level satellite data of SST supplied by NASA and historical tuna catch data supplied by SPC, using a geographical information system （GIS）, relations between the catch of tuna and SST were studied. With this information and using the Bayesian theory approach, a tuna fishing grounds forecasting expert system was set up and was developed generating probable fishing grounds charts. Bayes models are different from generic statistical method, in which not only the model information and data information, but also transcendental information is used adequately. The result of 40 years hindcasting experiments shows that the predicting accuracy of skipjack fishing grounds in West Pacific ocean is over 70 %, which is significant to guide fishermen＇ fishing operation. However, because the computation period of fishing grounds transcendental probability and conditional probability is every month, it must he modified according to field survey data for future fishing grounds forecasting ever week.